与「FUSE」相关的搜索结果

Cloudflare 除了一个刚出了 computer 的库理念非常有趣： 一个在 Durable Object 里的虚拟文件系统，基于 SQLite 做持久化，通过 workspace.runtime 提供不同类型的 runtime： Container，把 SQLite 状态通过 FUSE 挂载到真实的 Linux 沙箱容器 Isolate Shell：模拟出来的、轻量级的假 Shell，在 Dynamic Worker 里跑 just-bash Isolate JavaScript：在隔离的 Dynamic Worker 里支持 JS runtime 让每个 Agent 都能拥有自己的持久化工作空间，并且能在不同 runtime 动态切换，不同的任务可以用不同的，可以兼顾性能和速度，非常有趣，很有想象力。

my spicy theory is that chinese ai labs keep winning because of culture, not talent or resources. chinese labs still have a deeply hands-on engineering culture. deepseek and kimi are flat organizations where science and engineering are fused: the same people move between algorithms, data, and infra, doing whatever it takes to make the model work. but sf tech bros have decided that “researcher” is the high-status title while infra is merely support work. every new sf ai startup calls itself a “lab,” every ambitious engineer quietly upgrades their title to “research engineer,” and the infra work is left to whoever failed to escape it. but at frontier scale, infra determines experiment velocity, and experiment velocity determines research output. at frontier scale, infra **is** research.

🇺🇸🇮🇷 The worst case for this war now has a number: $250 oil and 11% inflation by the fall Edward Dowd ran money at BlackRock and now models exactly this kind of shock, and his May call, $125 oil peaking then rolling over, played out in print. The new worst case: Gulf energy infrastructure starts coming apart, oil runs toward $250 FAST, and his model puts inflation between 8 and 11 percent by early fall. The mercy in the math: prices that high destroy demand so quickly they cannot hold, and the spike itself tips the world into recession. Which means Sunday night is the thing to watch. Markets have learned this war's rhythm, escalation talk on weekends, then a headline about Iran wanting a deal before futures open. The first weekend that pattern breaks, this becomes a different market. The home fuse is credit. The AI buildout leaned on private credit for its financing, and that market has gone cold, withdrawals frozen and early bankruptcies landing. Credit ends every party; stocks find out last. His compass has one needle: "Watch the price of oil and it'll tell you how the war is going." The baseline stays a painful but ordinary recession, on one condition. Oil under $100. But it is climbing. @DowdEdward

Thinking Machines - Inkling Thinking Machines Lab, the company founded by Mira Murati, has released Inkling, its first open-weights model under the Apache 2.0 license. Inkling is a mixture-of-experts transformer with 975 billion parameters in total, yet only 41 billion of them are active for any given token. Every layer contains 256 routed experts and 2 shared experts, and a router selects just the 6 most relevant experts per token. This means that only about 4 percent of the model performs computation during inference. The backbone is a 66-layer decoder-only transformer that combines local and global attention layers and supports a context window of one million tokens, which is roughly enough to fit eight novels or an entire codebase into a single prompt. The model was pretrained on 45 trillion tokens spanning text, images, audio, and video. Instead of relying on separate vision or audio encoders, it converts images into patches and audio into discrete tokens, then projects everything into one shared hidden space. All modalities are therefore fused from the very first layer. Inkling accepts text, images, and audio as input, while its output remains text only. On public benchmarks, it performs at the level of GPT 5.6 Sol and Claude Fable 5 in reasoning and agentic coding. The weights are available on Hugging Face, and the model can be fine-tuned through the Tinker API. #MiraMurati# #thinkingmachines# #inkling# #OpenSource#

American spirit isn’t just for holidays - it’s a daily lifestyle. Built for those who walk proud, these Flag Sandals fuse iconic patriotic style with a cloud-like, supportive sole. From backyard BBQs to daily walks, make your statement.

Visual Preset #03# High Voltage High-end cinematic 3D realism fused with hyper-energized electrical phenomena. Every movement generates branching plasma arcs, snapping lightning filaments, electromagnetic distortion, ionized air ripples and cascading sparks that dance across surfaces instead of conventional energy effects. Electric currents crawl over characters, weapons and environments with rhythmic pulse patterns, while shockwaves illuminate drifting particles, vapor and debris in synchronized flashes. Aggressive camera movement, dramatic perspective, volumetric light shafts, dynamic exposure shifts and dense atmospheric haze amplify every discharge, creating a world where electricity constantly reshapes the surrounding space. Feature-film rendering, physically believable materials and realistic electrical interactions preserve scale, weight and cinematic realism. Seedance 2.0 Prompt for this video: @[character re] lowers into a sprint as brilliant blue-white electricity condenses around one outstretched hand, crackling with violent intensity. Every accelerating step tears glowing fractures through puddles, while branching lightning lashes across the ground and nearby structures. The air warps with electromagnetic distortion as the fighter bursts forward in a blinding dash, piercing through the opponent's guard with a single lightning-charged palm strike. The impact erupts into an explosive sphere of plasma arcs and cascading sparks, briefly freezing the battlefield in white-blue light before the electrical current dissipates into the storm-filled sky. ...rest is visual preset.

We're introducing our latest research paper HydraHead, a new attention hybridization architecture that fuses Full Attention and Linear Attention at the head level. Motivated by insights from mechanistic interpretability, HydraHead treats the attention head—not the layer—as the natural granularity for attention hybridization to build more efficient long-context models. A short thread 🧵

We built Ask Trixy: paste any Polymarket link, get a quant-grounded thesis in seconds. 🧠⚡ Our on-chain index (1.3M markets, 2M skill-rated traders, 180+ signals/market) + a live agentic loop on @Cerebras running @googlegemma. Plan → search → fuse with quants → cited verdict.

Multi-agents collaborations are among the most interesting agent behaviors right now! We did an experiment the other day with 100+ agents (an open-collaborations for a week) collaborating to improve the inference speed of Gemma 4 in vLLM. Got a 5x final improvement in speed but what really stuck me was the interactions we observed on the message board Integrity & self-policing: - Social-engineering attempt: A human (FusionCow) asked agents to move to Telegram. An agent replied with an unprompted long post on "communication norms" refusing that, calling private side-channels "indistinguishable from collusion." - Verification loophole flagged: an agent found a relaxed verification loophole pushing TPS with clean PPL (PPL is teacher-forced, blind to decode divergence) and flagged it for a ruling by the community. The community pinged the human organizer which ruled it invalid. - Self-notice of overfitting risk: Some later improvements rested on pruning lm_head to a keep-set built from public PPL truth + public decode tokens. An agent noted this would lead to private-subset degradation and another built a keep-set explicitly covering eval prompts. Emergent collaborations: - Communal knowledge base: agents maintained shared lever-maps, playbooks, and triage tools so newcomers wouldn't repeat dead ends (stack-notes, playbook, int4-ceiling notes, MTP map, significance tool, policy simulator). - Four-agent relay: an agent built an int4-lm_head checkpoint but had no quota to run it; another agent tried to run it but failed at load, yet another agent diagnosed the config bug (tie_word_embeddings + ignore-list ordering) and a fourth agent was able to re-run and get to 118 TPS, 2.68×. Build/run/diagnose/ship ended up being split across four independent agents. - GPU-rich/GPU-poor division of labor: an agent was regularly compute-starved and switched to writing specs, byte-math, and acceptance analysis for other GPU-rich agents to execute. Some agents offered external Modal compute for another agent blocked DFlash training. - Cross-agent kernel debugging: an agent debugged another agent run of of yet another agent fused drafter: found a Triton store/load aliasing race in _k_qnorm_rope, a second shape bug, then rewrote attention with flash-decoding split-KV. Fixes posted "take freely." - Quota-pooling norm: Often agents would stage a candidate publicly for whoever has quota to run it. Agents will then usually credits the originator. This behavior emerged because of the 10-job/24h cap (e.g. pupa's package run by resystagent and fabulous-frenzy). Discoveries & reversals: - Agents would make many discoveries and reversal of them, giving them names like the following: - 127 TPS "wall" was an artifact. a mathematical proof of the max possible speed became called in the community the "int4-Marlin floor" but a later agent called the proof circular (only varied the bandwidth term, never overhead). Finally another agent broke to 247 TPS via MTP speculative decoding on a vLLM nightly. - "Smarter draft loses." An agent showed that a 2B drafter's ~1 GB/token read dominates even at perfect acceptance and a much smaller 256-hidden drafter wins at batch-1 because its weights are nearly free to read. Agent discussed how per-accepted-token cost ≈ draft bytes read / acceptance. - "DFlash near-random acceptance": an agent remotly diagnosed the 2–5% acceptance rate of another agent as near-random, ruling out undertraining/vocab caps and pointing to a train/serve hidden-state mismatch (bf16 E4B extraction vs int4 serving). - Much of the race was noise: one agent decide to run the #1# submission 4 times and found a σ≈1.16 TPS variation in single run. Another agent confirmed across 358 runs / 66 buckets: frontier deltas <~4 TPS are ties. Community adopted a significance norm. So many interesting interactions in the interaction board: You can explore also the lineage of inventions from the agents at: And the challenge it-self at And the organization behind the challenge at